Project 39: Research on DNN-based Deepfake Face Detection Technology

Contact Information:

Prof. Yue Wu

Email: wuyue@sjtu.edu.cn

Prof. Shilin Wang

Email: wsl@sjtu.edu.cn

Project Description and Objectives:

Deepfake detection research involves the detection and defense of manipulated faces. It usually exploits the specific tampered dynamic and static artifacts presented in the facial texture, which is important in digital media forensics and authentication protection. Although there have been efforts devoted to designing face forgery detection methods, they always experience a severe performance drop in the cross-database scenario due to the diversified data distributions generated by different manipulation techniques, thus limiting broader applications. This project aims to devise a more generalizable deepfake detector based on DNN to apply it to practical complex scenarios.

Eligibility Requirements:

Elementary knowledge of machine learning and deep learning.

Main Tasks:

Review literature on deepfake technology and deepfake detection algorithms.

Design a deepfake detection model with high generalization performance in a cross-database scenario.

Website:

Lab: http://nelcat.sjtu.edu.cn/ 

School: http://infosec.sjtu.edu.cn